This model is a fine-tuned version of hZzy/mistral-7b-sft-25-1 on the hZzy/direction_right2 dataset. It achieves the following results on the evaluation set:

Loss: 0.4420
Objective: 0.4416
Reward Accuracy: 0.6655
Logp Accuracy: 0.6499
Log Diff Policy: 15.7908
Chosen Logps: -168.1699
Rejected Logps: -183.9607
Chosen Rewards: -0.7349
Rejected Rewards: -0.8890
Logits: -2.0174

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 5e-06
train_batch_size: 3
eval_batch_size: 3
seed: 42
distributed_type: multi-GPU
num_devices: 3
gradient_accumulation_steps: 12
total_train_batch_size: 108
total_eval_batch_size: 9
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: constant_with_warmup
lr_scheduler_warmup_ratio: 0.1
num_epochs: 1

Training results

Training Loss	Epoch	Step	Validation Loss	Objective	Reward Accuracy	Logp Accuracy	Log Diff Policy	Chosen Logps	Rejected Logps	Chosen Rewards	Rejected Rewards	Logits
0.6077	0.1517	100	0.5074	0.5043	0.5702	0.5254	0.9522	-93.2911	-94.2433	0.0139	0.0082	-2.2027
0.5413	0.3033	200	0.4873	0.4899	0.6037	0.5937	8.2778	-125.3062	-133.5840	-0.3062	-0.3852	-1.9290
0.5306	0.4550	300	0.4690	0.4694	0.6373	0.6267	12.8023	-128.2087	-141.0110	-0.3353	-0.4595	-1.9229
0.4809	0.6067	400	0.4566	0.4556	0.6507	0.6390	13.4130	-110.8082	-124.2212	-0.1613	-0.2916	-2.1010
0.5005	0.7583	500	0.4521	0.4527	0.6566	0.6538	16.5186	-141.8083	-158.3269	-0.4713	-0.6326	-2.0970
0.4543	0.9100	600	0.4501	0.4510	0.6692	0.6586	17.7658	-179.5024	-197.2682	-0.8482	-1.0220	-2.0605

Framework versions

PEFT 0.11.1
Transformers 4.42.0
Pytorch 2.6.0+cu124
Datasets 3.2.0
Tokenizers 0.19.1

hZzy/mistral-7b-expo-7b-L2EXPO-25-8

mistral-7b-expo-7b-L2EXPO-25-8

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

hZzy/
mistral-7b-expo-7b-L2EXPO-25-8